Squall wind plus a convective downpour.
Classified from the same drivers; Severe Local Storm and Tropical Cyclone compete directly on the wind term and are separated by the rainfall pattern and the other drivers.
Uncalibrated softmax, as above.
Nothing above the WATCH level can be published automatically while no calibration map exists, and nothing at all can be published while the pipeline stamps no model version — the alert engine records that as `publication_blocked` rather than issuing the alert.
Uses the peak 24-hour rainfall rather than the horizon total, precisely so a wet fortnight is not read as a squall line — the defect the corrected wiring removed.
Drivers: maximum wind in the horizon (W, km/h) · wettest 24 hours in that horizon (P_peak, mm).
clip(x) clamps a term to [0, 1] — max(0, min(1, x)) — which is how every formula in the pipeline bounds its terms. Each term is clipped before it is weighted, so no single driver can run away with a score.
The run this deployment ships (prediction date 2026-09-16) classifies no district-horizon unit as Severe Local Storm. An absence of severe local storm in one run is not a statement that the hazard cannot occur this season.
Typical season: March–May (Nor'wester / Kalbaishakhi), plus pre-monsoon squall lines. This describes when the hazard is climatologically plausible, not when this run flags it.
Ground-truth reports (district, hazard, horizon, date, what was observed) are the only route by which this project can publish skill metrics. The contact page reaches the maintainers.
No. The severity index is the model's continuous score for how strongly the drivers resemble this class, on a 0.00–1.00 scale. It is not a probability, not a percentage, and it does not model exposure. The physics cross-check is a second, independent estimate; where the two diverge, both numbers are shown rather than averaged.
The published confidence is the model's own softmax over its eight classes, labelled uncalibrated_model_softmax in the data. It says how certain the classifier is about the class it picked, not how often that class actually materialises. No calibration map has been fitted yet, which is also why nothing above the WATCH level can be published automatically.
The district pages cover all 64 districts. This page summarises the districts the current run covers — a run can be partial, and the coverage stamp on the snapshot says exactly how partial, which the status page reports.
Loading the interactive HazardNet application…